Real-time label-free detection of dividing cells by means of lensfree video-microscopy Srikanth Vinjimore Kesavan Fabrice P. Navarro Mathilde Menneteau Frederique Mittler Brigitte David-Watine Nelly Dubrulle Spencer L. Shorte Bernard Chalmond Jean-Marc Dinten Cedric P. Allier
Journal of Biomedical Optics 19(3), 036004 (March 2014)
Real-time label-free detection of dividing cells by means of lensfree video-microscopy Srikanth Vinjimore Kesavan,a Fabrice P. Navarro,a Mathilde Menneteau,a Frederique Mittler,a Brigitte David-Watine,b Nelly Dubrulle,b Spencer L. Shorte,b Bernard Chalmond,c,d Jean-Marc Dinten,a and Cedric P. Alliera,* a
Commissariat à l’énergie atomique et aux énergies alternatives (CEA), LETI, MINATEC, 17 rue des martyrs, Grenoble cedex 9, 38054 France Plateforme d’imagerie dynamique, Imagopole, Institut Pasteur, Paris, 75015 France c University of Cergy-Pontoise, UFR Sciences, Cergy 95011 France d CMLA, École Normale Supérieure de Cachan, 94230 France b
Abstract. Quantification of cell proliferation and monitoring its kinetics are essential in fields of research such as developmental biology, oncology, etc. Although several proliferation assays exist, monitoring cell proliferation kinetics remains challenging. We present a novel cell proliferation assay based on real-time monitoring of cell culture inside a standard incubator using a lensfree video-microscope, combined with automated detection of single cell divisions over a population of several thousand cells. Since the method is based on direct visualization of dividing cells, it is label-free, continuous, and not sample destructive. Kinetics of cell proliferation can be monitored from a few hours to several days. We compare our method to a standard assay, the EdU proliferation assay, and as proof of principle, we demonstrate concentration-dependent and time-dependent effect of actinomycin D—a cell proliferation inhibitor. © 2014 Society of Photo-Optical Instrumentation Engineers (SPIE) [DOI: 10.1117/1.JBO.19.3 .036004]
Keywords: lensfree video-microscopy; label-free cell proliferation assay; real-time cell culture monitoring; cell proliferation kinetics. Paper 130836R received Nov. 22, 2013; revised manuscript received Jan. 8, 2014; accepted for publication Jan. 23, 2014; published online Mar. 5, 2014.
1
Introduction
which overcomes the drawbacks of existing methods. Unlike the currently used proliferation assays, cell division is directly detected without the need for surrogate measurements, exogenous contrast agents, or fluorescent dyes. Further, it is practical and highly amenable to facilitate high throughput inasmuch as it (1) does not require cell harvesting and (2) provides continuous direct live imaging data to follow cell proliferation kinetics as they happen in thousands of cells, yielding robust statistics outright. To validate our methodology, we compare the results obtained from lensfree video proliferation assay and classical EdU proliferation assay. We further confirm the approach by testing its capacity to follow the changes in proliferation kinetics induced by treating the cells with a cell proliferation inhibitor.
Cell division is one of the main events determining cell fate. Cell proliferation rate can reveal important information on perturbation of the cell cycle and is used routinely in a variety of biomedical research areas, including oncology and drug discovery. Current cell proliferation assays quantify proliferation either directly by incorporating a modified nucleotide (BrdU/EdU) to the newly synthesized DNA (at S phase of cell cycle)1–3 or indirectly by measuring parameters such as total ATP/DNA content,4–6 metabolic rate,7,8 substrate impedance changes,9–12 etc. Most of the direct techniques are static end point assays. Hence, they do not allow the measurement of cell proliferation kinetics—a critical parameter to test the time-dependent effect of various drugs/agents on cell proliferation. Other limitations include dependency on markers, being cumbersome, and being sample destructive. Indirect techniques are unsatisfactory, as strong assumptions are needed to correlate surrogate measurements with single cell division. The simplest way to measure cell proliferation rate would be to count individual cell divisions in a cell population as and when it occurs. Only very few methods have been proposed so far to quantify/analyze cell division on the basis of time-lapse imaging.13–17 However, the limitations include limited field of view (FOV), high cost and decreased feasibility of the approach, phototoxicity, and photobleaching. Here, we propose a new method, coined “lensfree video proliferation assay,” based on continuous and high-throughput recording of cells in culture using lensfree video-microscopy (see Sec. 2). It features the automated detection of dividing cells among a population of thousands of cells at a glance,
2
*Address all correspondence to: Cédric P Allier, E-mail:
[email protected]
0091-3286/2014/$25.00 © 2014 SPIE
Journal of Biomedical Optics
2.1
Materials and Methods Lensfree Video-Microscopy
Lensfree video-microscopy mentioned here implements in-line holographic imaging technique, which is explored extensively of late.18–23 It consists of a 12-bit APTINA MT9P031 CMOS RGB imaging sensor with a pixel pitch of 2.2 μm, measuring 5.7 × 4.3 mm, and light-emitting diode (LED) (dominating wavelength 525 nm) with a pinhole of 150 μm. In a typical experiment, the lensfree video-microscope is placed inside the incubator and the petri dish containing the cells is placed on lensfree video-microscope. Illumination is provided by the LED along with the pinhole from a distance of ∼5 cm. The light scattered by the sample and the light passing directly from the source to the imaging sensor interfere to form
036004-1
March 2014
•
Vol. 19(3)
Kesavan et al.: Real-time label-free detection of dividing cells. . .
a holographic pattern, which is recorded by the sensor. A USB cord connecting the laptop and the system passes through the provision at the rear of the incubator (Fig. 1).
2.2
Cell Culture
Murine NIH3T3 fibroblasts were obtained from the American Type Culture Collection and were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% newborn calf serum (PAN Biotech, Aidenbach, Germany) and 1% antibiotics (penicillin and streptomycin) (Gibco, Life Technologies, Saint-Aubin, France). Primary cultures of human fibroblasts prepared from skin biopsies performed on healthy donors (male, 30 to 39 years) were established in DMEM-Glutamax (Invitrogen, Life Technologies, Saint-Aubin, France) supplemented with 20% fetal calf serum (FCS) and used at passage 2. Nemo−∕− (seeding density: 30,000 cells) cells are SV40 immortalized human fetal female fibroblasts with a mutant Nemo gene (a gift from Dr. A. Smahi). Vero cells (a gift from F. Barre-Sinoussi’s lab) (seeding density: 26,000 cells) are kidney cells from Cercopithecus aethiops. BJ newborn foreskin fibroblasts are from the American Type Culture Collection. Cells (Nemo, Vero, 343, BJ) were grown in DMEM Glutamax plus 10% FCS.
2.3
in the dark in presence of the click cocktail reaction buffer containing the fluorescent dye (Invitrogen). After washing and centrifugation at 500 g for 5 min, cells were harvested and stored at 4°C until fluorescence activated cell sorting (FACS) analysis.
Comparison of Lensfree Video Proliferation Assay with Standard EdU Proliferation Assay
A standard 35-mm petri dish was filled with 2.5 ml of culture media at a cell density of 2.5 × 104 cells∕ml. For cell incubation with EdU, cells were seeded on day 1 in DMEM supplemented culture media with 10% FCS. Cells were treated on day 2 with 10 μM EdU (Click iT® EdU Invitrogen) followed by 2.5-h incubation. The petri dish was imaged with lensfree videomicroscope during EdU incubation period. Following the incubation period, EdU was removed through three washes in phosphate-buffered saline (PBS) and cells were transferred in tubes after trypsin (Gibco) treatment. According to manufacturer’s recommendation, cells were fixed in 1 ml 1% formaldehyde (Sigma-Aldricha, L'Isle d'Abeau, Chesnes, Saint-Quentin Fallavier, France). After 15 min, cells were washed once in PBS/1% bovine serum albumin (BSA). Then, cells were treated for 30 min with 1× saponin solution for membrane permeabilization. After washing once in PBS containing 1% BSA, cells were centrifuged at 500 g for 5 min and supernatant was eliminated. Cells were incubated for 30 min at room temperature
2.4
Inhibiting Cell Proliferation Using Actinomycin D
A standard 35-mm petri dish was filled with 2.5 ml of culture media at a cell density of 2.5 × 104 cells∕ml. Cells were seeded in three petri dishes on day 1. On day 2, actinomycin D (ActD) was added at concentrations of 0 μg∕ml (untreated control), 5 μg∕ml, and 10 μg∕ml. The petri dishes were imaged simultaneously by three lensfree video-microscopes immediately following the administration of the drug for a period of 6.5 h. Similarly, three petri dishes were prepared in parallel adhering to the same protocol for EdU proliferation assay. EdU was added for the final 2.5 h without changing the concentration of ActD.
2.5
Monitoring Cell Proliferation Kinetics
A standard 35-mm petri dish was filled with 2.5 ml of culture media at a cell density of 2.5 × 104 cells∕ml. Cells were seeded on day 1. On day 2, cells were imaged using lensfree videomicroscope from ∼4.5 h before the addition of ActD (at 2.5, 1, and 0.5 μg∕ml) until ∼4.5 h following the addition of the drug.
2.6
FACS Analysis
EdU-incorporated cells were analyzed using a BD LSR II twolaser flow cytometer (BD Biosciences, San Jose, California). The red laser (633 nm) is used for the detection of Alexa Fluor® 647. Sample measurements were performed with DIVA® software (BD Biosciences). Cell debris and aggregates were excluded from the analysis using an appropriate threshold (∼30;000).
2.7
Computational Methods
For pattern recognition, a typical holographic pattern obtained from a dividing metaphase cell was chosen to act as a template. Using normalized cross-correlation function available in MATLAB® with a constant threshold, the template was matched with the full FOV image of 24 mm2 , and the cells exhibiting similar holographic patterns were recognized and counted.
Fig. 1 Schematic diagram (a) and photograph of lensfree video-microscope (b) that is used to obtain time-lapse images inside the incubator. Red arrows indicate lensfree video-microscopes placed inside a standard incubator (c).
Journal of Biomedical Optics
036004-2
March 2014
•
Vol. 19(3)
Kesavan et al.: Real-time label-free detection of dividing cells. . .
Fig. 2 Temporal images of interest from real-time monitoring of a BJ cell (Video 1) showing the change in cell shape and cell adhesion taking place during cell division. Cell rounding can be observed in images obtained at T0 þ 0 h 52 min and T0 þ 0 h 54 min. Reduced cell substrate adhesion leading to signature holographic pattern of a dividing cell can be seen in images T0 þ 0 h 56 min and T0 þ 1 h 18 min. Cytokinesis and separation of daughter cells can be seen from T0 þ 1 h 56 min until T0 þ 4 h 56 min. Diagonal (dotted) green line indicates the longest axis of the cell prior to cell division. Red line denotes the cell division axis during cytokinesis. Yellow line denotes the axis of separation of daughter cells. The cell in this case has divided along the longest axis prior to cell division. Scale bar: 100 μm.
To estimate the total number of cells present in an image (dividing and nondividing), the obtained image was converted to its binary form. Using constant appropriate thresholding based on gray-level, area, major, and minor axis lengths, holographic patterns corresponding to cells were identified and counted. Percentage of dividing cells is the ratio of total number of dividing cells to average total number of cells calculated from images obtained over a period of time (typically 2.5 h). A period of 2.5 h was chosen in order to be in concurrence with the EdU incubation period of 2.5 h. In cases of highly confluent cultures, the typical holographic signature corresponding to metaphase might be masked and left
undetected. In these cases, cell retraction was used as a way to detect dividing cells. Retracting cells exhibit a pattern where the center of the holographic pattern is bright, while the edges are dark (Fig. 2, images at T0 þ 0 h 52 min and T0 þ 0 h 54 min). By detecting these alternating bright and dark intensity regions in an image, retracting cells were identified (Video 1).
3 3.1
Results Real-Time Monitoring of Cell Culture and Automated Detection of Dividing Cells
Lensfree video-microscope works using the principle of in-line holography (Sec. 2). The absence of magnification lenses renders large FOV of 24 mm2 covering several thousand cells. Lensfree video-microscope is placed inside a standard incubator (as shown in Fig. 1) to monitor cell culture in real time. The automated detection of dividing cells using lensfree video-microscopy is based on the changes in shape and adhesion that mitotic cells undergo. Before dividing into two
Video 1 Changes in cell shape and cell adhesion during cell division from the perspective of lensfree video-microscope. Frames were obtained every 2 min for ∼5.5 h. Cell rounding and reduced cell-substrate adhesion during cell division gives a typical holographic pattern, which is considered as the signature for mitotic cells. Cell retraction, cell rounding, reduction in cell-substrate adhesion, cytokinesis, and separation of daughter cells of a BJ cell is observed. Scale bar: 100 μm. (mov, 419 KB) [URL: http://dx.doi.org/10.1117/1.JBO.19.3 .036004.1].
Journal of Biomedical Optics
Fig. 3 (a) Schematic drawing showing the change in cell shape during cell division typifying a process called mitotic cell rounding. (b) Cell rounding and reduction in cell-substrate adhesion preceding the separation of daughter cells is clearly observed in the hologram obtained from the NIH 3T3 cell at T ¼ T0 þ 20 min. The two daughter cells are seen at T ¼ T0 þ 100 min. Scale bar: 50 μm.
036004-3
March 2014
•
Vol. 19(3)
Kesavan et al.: Real-time label-free detection of dividing cells. . .
daughter cells, almost all mammalian cells undergo dramatic shape transformation from being flat during interphase to being round during M phase, typifying a process termed “mitotic cell rounding.” This is usually accompanied by a reduction in cell-substrate adhesion. A schematic drawing illustrating changes in cell shape and substrate adhesion during division is presented in Fig. 3(a). Upon entering mitosis, the complex actin network is completely deconstructed and re-formed.24 Hence, at metaphase, during mitosis, the cells adopt a round shape and a decreased adhesion to the substrate.25–29 This helps in efficient and stable bipolar spindle formation and is, thus, vital to ensure a proper cell division.30 Almost all proliferating animal cells undergo these changes before cytokinesis. While extensive research has aimed to understand the underlying mechanism of the driving force(s) leading to cell rounding,31–36 these changes have not been exploited as a signature of mitotic cells. By contrast, lensfree video proliferation assay exploits the change in cell shape and cell adhesion as a natural marker to detect dividing cells. Jin et al. also mentioned the changes in the holographic pattern corresponding to the changes in cell morphology during cell division,22 but it was not exhaustively studied. It is to be noted that the holographic pattern obtained from a floating cell is different from the holographic pattern obtained from an adherent cell. When the cell is not adhered to the substrate, the holographic pattern obtained from the cell is similar to an airy pattern. The zero-order gray-level value is lower and interference rings are observed. In contrast, for an adherent cell, the zero-order gray level reaches larger values and the holographic pattern obtained is different. These changes can, hence, be observed during cell division as there is reduction in cellsubstrate adhesion. To exemplify the approach, a montage of the recorded holographic pattern of a dividing NIH3T3 cell observed using the lensfree video-microscope is shown in Fig. 3(b). At T ¼ T0, the cell is elongated and adhered to the substrate with larger zero-order gray-level value. There is a sharp change in the holographic pattern obtained from the
cell at T ¼ T0 þ 20 min. The zero-order gray level reaches a lower value denoting reduction in cell-substrate adhesion. The daughter cells are observed at T ¼ T0 þ 100 min. All the cells that experience rounding and reduction in cell-substrate adhesion during division exhibit a similar holographic pattern [as in Fig. 3(b), T ¼ T0 þ 20 min]. Thereafter, cells that are in the process of division are identified among several thousand neighboring cells by pattern recognition (FOV of the image 24 mm2 ) [Fig. 4(a)]. In some cases, the pattern corresponding to cell division may appear in subsequent temporal images. These repetitive events, though not counted, are marked by yellow squares [Figs. 4(a), 4(b), and 4(c)]. We have verified the accuracy of the automated detection of patterns by manually detecting the patterns and by following it in the subsequent temporal images to ensure the occurrence of cell division. Since true negatives cannot be determined in this case, we calculated the F1 measure (harmonic mean of precision and recall with equal weightage) based on true positives, false positives, and false negatives. True positives constitute accurate automated detection of patterns from cells that further divide in the subsequent temporal images. False positives constitute erroneous detection of patterns, which either were not from cells or were from cells that did not divide in the subsequent temporal images. False negatives are the cell divisions that were missed. The F1 measure is close to 0.87 on a scale of 0 to 1, with 1 being the best score. The measurement was based on seven random sequences of nine images that are temporally separated by 20 min, of dimension 1900 × 1425 μm, from different independent experiments (Table 1).
3.2
Comparison of the Lensfree Video Proliferation Assay with Standard EdU Proliferation Assay
In order to validate the method, it was directly compared to the standard EdU proliferation assay. Cells were imaged using the lensfree video-microscope inside an incubator during the EdU incubation period of 2.5 h. The acquired images were subjected
Fig. 4 (a) Number of cells exhibiting the pattern corresponding to cell division is detected using pattern recognition in a full field of view image of 24 mm2 , spanning across several thousand cells. New cell divisions are encircled in green, while repetitive cell divisions (cells that were round from the previous image) are marked in yellow squares. (b) Magnified region of interest showing cells that are encircled in green and marked in yellow squares among adherent (bright holographic pattern) neighboring cells. Scale bar: 50 μm. (c) Time-lapse images showing a cell exhibiting the pattern corresponding to cell division in subsequent temporal images (temporally separated by 20 min). Initially, the cell is encircled in green and later it is marked in yellow square. Scale bar: 50 μm. Journal of Biomedical Optics
036004-4
March 2014
•
Vol. 19(3)
Kesavan et al.: Real-time label-free detection of dividing cells. . .
to pattern recognition. A total of nine images were acquired per experiment. The number of cells undergoing division was calculated for each image and was summed. In order to compare the results obtained with the two assays, the total number of dividing cells calculated with the lensfree video proliferation assay was divided by the average total number of cells that were present in the images to yield the percentage of dividing cells. Automated detection of the total number of cells (in dividing and nondividing cells) was based on binary conversion of the image and an appropriate threshold to remove noise. The number of cells from random images was manually calculated to verify the accuracy of automated detection (Table 1). Six independent experiments were performed. It was observed [Fig. 5(a)] that the proliferation rate (percentage of dividing cells) obtained using our method is 18 5% (s.d. n ¼ 6 experiments), whereas the proliferation rate obtained using EdU proliferation assay is 33.8 6% (for the EdU incubation period of 2.5 h). The difference in the results could be due to false positives from EdU proliferation assay or false negatives from lensfree video proliferation assay (or both). EdU proliferation assay detects the cells that are in the initial stage of cell division (S phase). False positives may occur in EdU proliferation assay if a cell that is marked during the S phase is stopped from dividing at the G2-M checkpoint [Fig. 5(b)] due to improper replication of DNA. Indeed the cells are allowed to pass through the checkpoints only after they have repaired DNA damages.37–39 Lensfree video proliferation assay detects the cells that are in the final process of cell division (M phase) and is unbiased by the G2-M checkpoint. However, false negatives may occur if holographic signature corresponding to mitosis is not well detected. Also, the temporal resolution of the experiments performed was 20 min; this may have also limited the detection efficiency causing false negatives. Nevertheless, the standard deviation of the proliferation rate measurement over N ¼ 6 experiments is 5%, showing the consistency of the method comparable to that of EdU proliferation assay. Further, the impact of a cell proliferation inhibitor is clearly depicted in the results obtained by the method (Sec. 3.3). Though we cannot claim that the method is quantitative at this stage, it is qualitative and can monitor the differences in cell proliferation, which is essential in various studies, especially drug/siRNA screening.
3.3
Inhibiting Cell Proliferation Using Actinomycin D
The rate of cell proliferation may be altered by various stimulating or inhibiting conditions or agents. We further assessed our methods by measuring the influence of ActD, which is well known for inhibiting cell proliferation.40,41 ActD was added to culture plates at predetermined concentrations of 0 μg∕ml (control), 5 μg∕ml, and 10 μg∕ml. Following the administration of the drug, the culture plates were imaged in parallel using three lensfree video-microscopes (as shown in Fig. 1) for 6.5 h. Images were acquired every 20 min, and the obtained images were subjected to pattern recognition to calculate the number of dividing cells. It is noteworthy that manipulation of the culture plates during the addition of the drug triggered the detachment of a few cells that also gave rise to a holographic pattern similar to the one corresponding to cell division. In order to avoid the interference of these floating cells in the calculation, the initial three images following the addition of the drug were not considered for measurement. Journal of Biomedical Optics
The number of dividing cells was calculated for a total of 324 images obtained from six independent experiments, with 108 images per condition (control, 5 μg∕ml, and 10 μg∕ml). As shown in the graphs (Fig. 6), the number of dividing cells was, on an average, between 30 and 40 per image for untreated cells, and a total of 625 66 (s.d. n ¼ 6 experiments) cells divided during the experiment time frame. On the contrary, the number of dividing cells was reduced to
